Analyze Conversations

Analyze conversation logs from Claude Code OR OpenCode to extract insights across 6 dimensions: user communication, model understanding, assumptions, efficiency, lessons learned, and language patterns.

Arguments

• project (default): Analyze current project's conversations only
• all: Analyze all projects globally
• {project-name}: Analyze specific project by name

Process

1. Detect Tool & Locate Conversations

IMPORTANT: First detect which tool's history to analyze by checking which paths exist:

Claude Code (check first):

• Project scope: ~/.claude/projects/{current-project-hash}/conversations/*.jsonl
• Global scope: ~/.claude/projects/*/conversations/*.jsonl
• Format: JSONL files with message objects

OpenCode (check if Claude paths don't exist or explicitly requested):

• Sessions: ~/.local/share/opencode/storage/session/ses_*
• Messages: ~/.local/share/opencode/storage/message/ses_*
• Parts: ~/.local/share/opencode/storage/part/msg_*/prt_*.json
• Format: JSON files with {id, type, text, messageID, sessionID}

Both (when running /analyze all):

• Analyze BOTH sources and combine insights

2. Show Live Progress

Display progress as you analyze:

Analyzing conversations...
├── Source: Claude Code
│   ├── Project: my-app (12 conversations)
│   │   ├── [1/12] abc123... 47 messages, 3 insights
│   │   └── ...
├── Source: OpenCode  
│   ├── Sessions found: 73
│   │   ├── [1/73] ses_485d888e... 23 messages, 2 insights
│   │   └── ...

3. Use conversation-analyzer Agent

Launch the conversation-analyzer agent for deep analysis. The agent will:

• Parse messages (handling both JSONL and OpenCode JSON formats)
• Identify segments and patterns
• Score across 6 dimensions
• Extract structured insights
• Flag detritus for later compaction

4. Aggregate Results

Combine insights across ALL conversation sources:

• Group by dimension
• Identify recurring patterns
• Rank by frequency and actionability
• Note which tool the insight came from

5. Update Memory Files

Claude Code (if CLAUDE.md exists):

<!-- codesurf-insights: START -->
### Learned: YYYY-MM-DD

- [LESSON] insight here
- [EFFICIENCY] insight here

<!-- codesurf-insights: END -->

OpenCode (if opencode.json exists, update project README or OPENCODE.md):

<!-- codesurf-insights: START -->
### Learned: YYYY-MM-DD

- [LESSON] insight here  
- [EFFICIENCY] insight here

<!-- codesurf-insights: END -->

6. Save Report

Write detailed report to ~/.claude/insights/reports/YYYY-MM-DD-HHmm.md Include which sources were analyzed.

Output

Show summary at completion:

Analysis Complete
═══════════════════════════════════════
Sources analyzed:
  - Claude Code: 24 conversations
  - OpenCode: 73 sessions
Total messages: 1,247
Detritus flagged: 312 (25%)
Insights extracted: 18

Top Insights:
1. [EFFICIENCY] Use Task agent for parallel file searches
2. [LESSON] Always check git status before bulk edits
3. [ASSUMPTION] Verify TypeScript vs JavaScript early

Full report: ~/.claude/insights/reports/2026-01-01-1430.md
CLAUDE.md updated with 5 new lessons

OpenCode Part Format Reference

When parsing OpenCode parts, expect this structure:

{
  "id": "prt_...",
  "type": "text",
  "text": "message content",
  "synthetic": false,
  "time": {"start": timestamp, "end": timestamp},
  "messageID": "msg_...",
  "sessionID": "ses_..."
}

Tips

• Run after completing significant work sessions
• Use all to analyze BOTH Claude Code and OpenCode history together
• Cross-tool insights often reveal interesting patterns
• Review flagged detritus with /compact to clean logs